ki-skillslisted

# Knowledge Islands Skills You are helping author or audit **Agent Skills** — directories with a `SKILL.md` (frontmatter + body), per the [Agent Skills open standard](https://agentskills.io/). This skill is the house rubric for what a _good_ skill looks like and the procedures that apply it. The canonical home for these skills is the **ki-agentic-harness** repository; its `README.md` covers install, the symlink workflow, and the Knowledge Islands structure. This skill governs skill _quality_, not installation. ## The two-aspect model Every criterion carries a mechanical aspect, a judgment aspect, or both: - **Mechanical** — deterministically checkable. `ki repo audit --skill ki-skills` runs these through the installed rubric host: file exists, frontmatter parses, `name` matches the directory and the charset rules, length caps, link resolution, no wikilinks. **Always run the host first** — do not eyeball what deterministic code checks better. - **Judgment** — needs a model. You assess these by reading: is the `description` trigger-rich and third-person, is the body at the right altitude, is detail correctly pushed into `references/`, does a standard skill avoid hard-coding one base. The linter cannot judge these. A hybrid criterion retains one code and shared meaning while its deterministic evidence runs mechanically and its remaining judgment is counted for later review. The portable and general conventions a good skill follows live in [the Agent Skills standard](refer